On Fri, Dec 20, 2019 at 05:32:55PM -0800, John Hubbard wrote: > Hi, > > This came up when I was running rdma-core tests on a two-machine setup, > where each card had two ports, but there was only one cable. So only > one port on each end was connected. > > The main thing I expect to be up for debate is, what string to return > for speed, when a port is disabled or down? I initially thought about > returning '(Disabled/down)', but it seems more accurate to just report > '0.0 Gbps', so that's what I settled on. > > Background: here's what I wrote when discussing this over on linux-mm > with Leon [1]: > > It looks like this test suite assumes that every link is connected! > (Probably in most test systems, they are.) I don't remember whenever the expectation of connection is by design or outcome of mine and Jason's setups, where our cards are being connected in loopback mode (port 0 to port 1 of the same card). The loopback mode simplifies our kernel testing and development. Thanks
